Claude as your RevOps system: the setup guide
How to set up Claude as the revenue operations system of your company: strategy first, one neutral data store, a semantic layer, clean data, then automation, Slack and the CRM question.
- 01Strategy and guardrailsThe brain: what good looks like
- 02One neutral databaseAll company data, owned by you
- 03Semantic layerOne definition for every number
- 04Clean data, kept cleanGarbage in, garbage out
- 05Claude, aiming for L3+From drafting to operating
- 06Scheduled workTasks and workflows that run themselves
- 07Slack and the CRMWhere people work, and what the CRM is for
“The first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency.”
Let’s go back to Berlin-Mitte. MyGym24, the fictional fitness software company I used throughout The RevOps Pendulum, has grown up. It now sells to gym chains across Germany, Austria and Switzerland, and the leadership team has one clear ambition: keep growing without doubling the team. So they did what many companies did last year. They gave everyone a Claude licence.
Did it work? Partly. Six months later, people write emails faster, the marketing team produces more content, and the sales reps prepare their calls in half the time. But the sales process, the board pack and the Monday pipeline call look exactly the same as before. Every person got a little faster. The company didn’t.
That is the difference between using AI and running your company on it. In the first case, Claude helps one person with one task. In the second, Claude can see the whole business, understands what every number means, and can act on what it finds. The second one doesn’t come from a clever prompt. It comes from groundwork, done in the right order.
This guide walks you through that order, layer by layer.
Most teams use Claude one question at a time. The real gain comes when Claude can see the whole business, knows what every number means and can act on what it finds. That takes groundwork, in this order.
- Strategy before operations. Claude needs to know what good looks like for your business, and someone has to keep that knowledge current.
- One neutral home for all your data. Commercial, finance, legal, product and people data in a database you own. Not the CRM.
- A semantic layer on top. One written definition for every metric and key process, so the same question gets the same answer in every tool.
- Clean data, kept clean. Garbage in, garbage out. AI just produces the garbage faster.
- Then connect Claude, and aim for automation at L3 and above, where most of the value sits.
- Work where your team works (Slack), decide what is left for the CRM, and keep control of cost.
Using Five Dots? Tailor this guide first
If you have the Five Dots connector in Claude, copy this prompt into a new conversation before you read on. Claude uses the Five Dots methodology and your own workspace to turn the guide into a plan for your business. Nothing about your company is sent to Five Dots.
I'm working through the Five Dots guide "Claude as your RevOps system". Use Five Dots to tailor it to our company.
1. Read our Five Dots workspace first (Memory, Decisions, earlier audit results) so you don't ask me things you already know.
2. Then go through the guide's layers in order: strategy and guardrails, a neutral data store, the semantic layer, data quality and governance, Claude and automation (L1 to L5), scheduled work, collaboration in Slack, and the CRM and cost question.
3. Follow the Five Dots knowledge article "Claude as your RevOps system" step by step. For each layer, tell me where we stand today, what is missing, and the next one to three actions. Ask me only for what you can't find.
4. Finish with a 90-day sequence and write it to our workspace as a Decision draft for me to approve.No Five Dots account yet? Create a free one and run the audit as well. No paid account is required.
What you get when it works
Before we start building, let me show you where we are going. Back to MyGym24, a year later, with the system in place. Pick a moment:
“Why did conversion jump in Austria, and can we repeat it?”
AustriaGermanySwitzerlandDemo-to-paid conversion, weeks 1–12
- 1Week 3: Price changeAll countries
- 2Week 5: New account executive in ViennaAustria only
- 3Week 7: Nutrition add-on launchAll countries
- 4Week 9: Partner newsletterAustria only
Claude: Only two changes were Austria-only. The jump starts in week 9, the week of the partner newsletter; the new account executive started four weeks earlier. Worth testing in Switzerland: a Swiss partner newsletter.
That is the promise: a lean commercial organisation that can do a great deal with very few people.
Read the other 10 sections, free.
Enter your email and the rest of the guide opens on this page: the setup, layer by layer, with the examples, the tables and the interactive figures. We’ll also email you a link so you can open it on any device.
- 03Strategy comes before operations
- 04Give all your data one neutral home
- 05Add a semantic layer: one definition for every number
- 06Clean your data, then keep it clean
- 07Now connect Claude, and aim higher than drafting
- 08Run it on a schedule
- 09Work where your team already works: Slack
- 10Decide what the CRM is still for
- 11Keep control of cost
- 12Where to start
Sources & further reading
- Claude scheduled tasks (cloud), Claude Code documentation
- Cowork scheduled tasks, Claude Help Center
- Set up Claude Tag, Claude documentation
- Give Claude Tag access to your tools
- Connect a service that isn’t in the list (Claude Tag)
- n8n MCP server and nodes
- n8n pricing
- Supabase MCP server
- BigQuery MCP server
- Snowflake-managed MCP server
- Fabric Data Warehouse MCP server
- dbt MCP server tools
- Cube MCP server
- dbt Semantic Layer documentation
- What is a semantic layer? (Cube)
- Henry Shi on joining Anthropic
- The Autonomous Business Scale, Growth After Headcount
- Salesforce Agentforce pricing
- Salesforce Agentforce rate card
See where you stand on each layer.
Run the Five Dots audit in your own Claude. Your data stays in your workspace.